par Zakir, Raina 
;Salahshour, Mohammad;Dorigo, Marco ;Reina, Andreagiovanni
Référence Lecture notes in computer science, 14987 LNCS, page (112-126)
Publication Publié, 2024-11-01
;Salahshour, Mohammad;Dorigo, Marco ;Reina, Andreagiovanni Référence Lecture notes in computer science, 14987 LNCS, page (112-126)
Publication Publié, 2024-11-01
Article révisé par les pairs
| Résumé : | We study how robot swarms can collectively adapt to dynamic environments by changing what they collectively select as the best among a set of n possible options. While the robots rely on local communication with one another, follow simple rules, and make estimates of the option’s qualities subject to measurement errors, the swarm as a whole can infer the change in the environment to make accurate collective decisions. Most studies focusing on dynamic environments have achieved adaptive behaviour by including random noise or threshold-based approaches to continuously explore alternatives and prevent opinion stagnation once consensus is achieved. In this study, we investigate whether or not swarms of robots with heterogeneous behaviours can be more adaptive than homogeneous swarms. We consider two behaviours from the literature which robots use to update their opinions: the majority rule, where robots gather information from all neighbours, and the voter rule, where robots use information from a single neighbour. In static environments, swarms of majority-rule robots, by using a larger amount of social information, typically make quicker decisions than swarms of voter-rule robots. However, our multiagent and robot simulations show that including voter-rule robots within a swarm of majority-rule robots can increase the group’s responsiveness to environmental changes. This result shows the potential benefits of mixing simpler and relatively more advanced robots in the same swarm. |
